Data Scientist — Refine Student and Loan Risk Scoring Model

Our Foundation’s existing student and loan-risk scoring model.

Our team have started developing the model with AI, but its score definitions, variables, assumptions, formulas, weights, and thresholds still need to be reviewed by an experienced data scientist.

The volunteer will define what the proposed “AI risk score” and “loan risk score” should measure, determine which student and education-related data should influence the scores, and refine the scoring methodology. Factors may include tuition, scholarships, grants, remaining net cost, the student’s circumstances, selected school or program, and comparable educational options.

The volunteer will test the model using realistic sample cases, including Laurens’s master’s-degree example, and identify inaccurate, unfair, biased, or misleading results.

Expected deliverables include a model assessment, refined scoring methodology, test-case workbook, data dictionary, fairness and sensitivity review, documentation of limitations and confidence levels, recommendations for future data collection, and clear implementation requirements.

Website, Salesforce, NetSuite, and intake-workflow integrations will be handled separately by an AI workflow and automation specialist after PYT approves the scoring methodology. The model should support human decision-making and should not automatically approve or deny students.

Through this project, the volunteer will help PYT make more consistent, transparent, and evidence-informed decisions for students facing last-mile education funding gaps.

The refined model will help PYT evaluate tuition, scholarships, remaining net cost, student circumstances, and comparable educational options while clearly explaining the reasoning and limitations behind each score.

The volunteer’s work will also help protect students from unfair or misleading outcomes and give PYT a responsible foundation for improving the model as more student, repayment, completion, and employment data becomes available.

The final methodology will also provide a clear handoff to the automation specialist who will later deploy the approved model across PYT’s systems.

PYT’s leadership supports this project and understands that the scoring methodology must be developed and validated by a qualified data scientist before it is deployed.

We will provide the volunteer with the current model, formulas, prompts, variables, assumptions, sample outputs and available historical data. We will also designate a primary project contact, schedule regular working sessions, and make leadership available to answer questions and approve key decisions.

Missing information or data limitations will be clearly documented rather than filled with unsupported assumptions. After the methodology is finalized, a separate automation specialist will use the volunteer’s implementation requirements to connect the model to PYT’s website, Salesforce, NetSuite, and intake workflows.